[Transformation of Epidemiology in the Age of Artificial Intelligence]

Epidemiology has been fundamental for analyzing health problems and supporting decision-making in healthcare systems and public health. However, traditional epidemiological methods, designed fundamentally to identify causal associations at the population level through aggregate data measures, present inherent limitations in capturing individual heterogeneity in response to specific exposures. This population-based approach hinders personalized prediction of outcomes in a particular individual whose risk factors may manifest differently from the group average, particularly when multiple contextual variables and unique biological profiles are involved. Advances in artificial intelligence have generated tools capable of integrating large volumes of information, identifying complex patterns in specific subgroups, and producing more personalized estimates, transitioning from a reactive approach based on population averages toward predictive models centered on individual trajectories. However, these developments do not replace the methodological foundations of epidemiology, as the identification of exposures, outcomes, and causal relationships continues to depend on the epidemiological conceptual framework. From this perspective, current tensions do not represent a disciplinary crisis, but rather a transition toward broader approaches that combine population-based analyses with advanced predictive tools. This integration is particularly relevant for large-scale healthcare institutions and national health systems, which require models capable of leveraging diverse data to improve understanding of health processes and support clinical and operational decisions.